Originally, we set out to make “meetings more efficient” by helping companies hold video conferences asynchronously. A crucial component in making that possible relied heavily on ASR (Automatic Speech Recognition) that would allow users to gather important highlights from a conversation via text. Many API services such as Google Speech API, Amazon Transcribe, and others boast their Speech-to-Text APIs have a 90% level of accuracy, we thought this would be a simple integration. We were wrong. We thought 90% accuracy was “good enough“. It was not.
The “10%” of words missed happen to also be the most crucial words of a sentence. Missing those words meant no usable information could be derived from the data. Since Anders is from Denmark, we tested how well the leading services fared in his native tongue. The results were terrible. We, and the industry, had a problem and had to rethink our mission.
We soon found out we weren’t the only ones experiencing the reality around the usability of ASR technologies. We cut out the meeting technology and decided to focus completely on allowing users to get great results without the need to code, transcode, or overcome all the other infrastructural challenges we experienced while building our meeting product.
As we dove headfirst into the problem, we discovered more major issues with current offerings in the market. The industry boasted many scientists, but few user-first, user-experience focused products. Additionally, no one made it their their mission to ensure that content provided by the user was securely stored and encrypted, compliant with GDPR standards, or handled with the user’s privacy in mind.
Konch has since become the very thing we, and others, were missing. An insanely accurate, incredibly fast, thoughtfully private Human Augmented AI platform. We have gone head to head with every competing service out there, and, when compared, have yet to lose.